robo-dimag: Giving AI Agents One Memory They Can Trust

This document is a plain-language companion to the robo-cortex paper (DOI 10.5281/zenodo.21610870). It tells the story of what came after: how a small memory store for one AI coding agent grew into robo-dimag, a federated memory gateway used daily by several agents across two servers and two VPS. The document covers seven concrete problems encountered between August and November 2026 — from removing a redundant search layer after measurement, to solving silent staleness in working notes — and how each was resolved, usually by taking something away rather than adding it. It includes two months of real usage telemetry (1,984 recall calls, median 1,857 tokens returned, 2,473 writes) and an honest estimate of token savings versus counterfactual lookup cost. Written for practitioners building memory systems for AI agents. All numbers come from the project's own logs and changelog.

Authors

Publication Details

Journal
Zenodo (CERN European Organization for Nuclear Research)
Published
2026-11-13
DOI
https://doi.org/10.5281/zenodo.24207415
Primary Topic
Software System Performance and Reliability
Type
preprint
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
preprint

robo-dimag: Giving AI Agents One Memory They Can Trust

Sorin Neagu, Sorin Robotel
2 links
Zenodo (CERN European Organization for Nuclear Research)
Software System Performance and Reliability
preprint

robo-dimag: Giving AI Agents One Memory They Can Trust

Sorin Neagu, Sorin Robotel
preprint en

Abstract

This document is a plain-language companion to the robo-cortex paper (DOI 10.5281/zenodo.21610870). It tells the story of what came after: how a small memory store for one AI coding agent grew into robo-dimag, a federated memory gateway used daily by several agents across two servers and two VPS. The document covers seven concrete problems encountered between August and November 2026 — from removing a redundant search layer after measurement, to solving silent staleness in working notes — and how each was resolved, usually by taking something away rather than adding it. It includes two months of real usage telemetry (1,984 recall calls, median 1,857 tokens returned, 2,473 writes) and an honest estimate of token savings versus counterfactual lookup cost. Written for practitioners building memory systems for AI agents. All numbers come from the project's own logs and changelog.

Zenodo (CERN European Organization for Nuclear Research)
Software System Performance and Reliability
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon